Thanks for the response KLiK but that was not what I was asking. I'll try to rephrase my current issue. I am going through these issues because when I initially installed this client my system was moving through work units at an approx time of 24 units per 3 hours and has slowed down drastically. I do not recall the performance of my original configuration since I was not monitoring it but as you can see something has changed and I cannot figure out what it is.

9/3 is when I first joined WCG
9/4 is when I had my system crunching for my first full day.
9/11 is when I did a fresh install of windows 7.
You will notice the points decrease from there.

Statistics Date Total Run Time Points Generated Results Returned
(y:d:h:m:s)
09/11/2016 0:001:11:59:07 2,469 6
09/10/2016 0:033:03:26:10 47,867 115
09/09/2016 0:034:20:29:02 52,117 124
09/08/2016 0:033:16:22:59 48,846 117
09/07/2016 0:032:21:32:51 59,671 126
09/06/2016 0:024:01:01:40 68,074 143
09/05/2016 0:025:00:03:48 100,922 246
09/04/2016 0:037:04:03:56 142,477 328
09/03/2016 0:006:17:58:19 28,445 77

9/5 Scenario 1: Computer is running Windows 10, Installed WCG Client (Latest), All compute settings set with no restriction. CPU Usage does not exceed 80%, hovers around 77%. Each core has inconsistent load. Approx time to complete a batch of work unit 7 hours. Work units being worked on at once 25-30.

9/7 Scenario 2: Computer is running Windows 10, Installed BOINC Client (Latest) from Berkeley website, All compute settings set with no restriction. CPU Usage is at a constant 100%. Each core has full load. This time however each batch is now 14-20 hours. Work units being worked on at once is 45-50.

9/11 Scenario 3: Got fed up with the inconsistency, Fresh install Windows 7, shut down all non essential services for minimum OS load. Installed WCG Client (Latest). All compute settings non restrictive, 10GB hdd space, 0.00 multiprocessor, 1 day min work que, 100% usage. CPU usage hovering around 70%, Time to complete a work unit 6-8 hours (still taking notes on this). Work units being worked on at once 26.

Hey Joe,

After you pointed me to the log I noticed that with the version downloaded fromWCG it was forcing the max cpu count to 32. However even when I managed to get it up to the full 48 with the version located on the Berkeley website it was running even slower.

Anyways i found out the problem was HyperThreading. After I disabled it all 24 physical cores handle their own work unit in about 2 hours.

Marty@MM,

From close observation, you can just set your client to 50% of processors **, which then allows 24 of your 48 threads, at which point the OS does not use hypertheading i.e. you would still have benefit of the feature for other purposes than crunching.

** Alternate, there's the <ncpus>n<ncpus> config option in cc_config.xml to limit the client. .A third way is to set a limit is in app_config.xml, telling WCG specifically not to run more than NYou can even give it a limit at app level with <max_concurrent>1</max_concurrent>

The manual: http://boinc.berkeley.edu/wiki/Client_configuration

Some large device owners have reported a performance issue with [IIRC] HSTb when going over 32 threads on a 64 threaded system... something gets saturated, but not heard of this being an issue for the other science.

Snip
The first difference is the number of validations assumed to be due PV Jail as we refer to ** ... New 134 v Old 328. At 2 hours, those would be guesstimated as being OET/ZIKA maybe UGM (They run regular and an hour longer though). 324*38,667/134 gives about 95K points. Delays in validation is suggested. The Time stat also indicates same 13:10 days v 37:04, so would suggest to let it rip a few days to stabilize the validation flow, where 24 cores at 24/24 would eventually balance at producing... ~24 days of runtime daily.

Keep in mind that though hyperthreading slows the results, there's a general uplift at the end of about 15 percent or more production... not necessarily in points, which is an extremely dodgy measure with the broken credit logic in place (search forums for discussion, many).

** Visit My Contribution > Result Status to review your returns.